Earlier work reported in the STL‐QPSR (Speech Transmission Laboratory, Quarterly Progress and Status Report, KTH, Stockholm) each year since 1979 has been extended to the modeling of the supraglottal impedance to include five formants and a time span of analysis to embrace several successive glottal periods. The acoustic interaction is highly nonlinear and may give rise to spectral peaks between formants. This is largely due to nonlinear superposition ripple originating from previous excitations. At a high F0, coinciding with F1, the air consumption is minimized as previously shown experimentally by Rothenberg. One aspect of the interaction is intraglottal variations in formant bandwidths and frequencies. The modeling of superposition at increasing F0 and constant articulation shows how the formant amplitude range at in‐phase and out‐of‐phase superposition decreases due to intraglottal losses. In addition, there appear nonlinear deviations such as a weak tendency of synchrony between the second and first formant amplitude dependency of F0. A constant leak added to the normal glottal area function has been modeled as a first step toward the study of leaky voices. Here, the apparent linear‐type formant ripple in the maximally closed phase is a memento for inverse‐filtering criteria. Implications for female and child phonations are discussed.

This content is only available via PDF.